Evolutionary Paths of the camp-dependent Protein Kinase (PKA) Catalytic Subunits

Transcription

1 Evolutionary Paths of the camp-dependent Protein Kinase (PKA) Catalytic Subunits Kristoffer Søberg 1,2, Tore Jahnsen 2, Torbjørn Rognes 3,4, Bjørn S. Skålhegg 1, Jon K. Laerdahl 4,5 * 1 Department of Nutrition, Institute of Basic Medical Sciences, University of Oslo, Oslo, Norway, 2 Department of Biochemistry, Institute of Basic Medical Sciences, University of Oslo, Oslo, Norway, 3 Department of Informatics, University of Oslo, Oslo, Norway, 4 Centre for Molecular Biology and Neuroscience (CMBN), Department of Microbiology, Oslo University Hospital Rikshospitalet, Oslo, Norway, 5 Bioinformatics Core Facility, Department of Informatics, University of Oslo, Oslo, Norway Abstract 39,59-cyclic adenosine monophosphate (camp) dependent protein kinase or protein kinase A (PKA) has served as a prototype for the large family of protein kinases that are crucially important for signal transduction in eukaryotic cells. The PKA catalytic subunits Ca and Cb, encoded by the two genes PRKACA and PRKACB, respectively, are among the best understood and characterized human kinases. Here we have studied the evolution of this gene family in chordates, arthropods, mollusks and other animals employing probabilistic methods and show that Ca and Cb arose by duplication of an ancestral PKA catalytic subunit in a common ancestor of vertebrates. The two genes have subsequently been duplicated in teleost fishes. The evolution of the PRKACG retroposon in simians was also investigated. Although the degree of sequence conservation in the PKA Ca/Cb kinase family is exceptionally high, a small set of signature residues defining Ca and Cb subfamilies were identified. These conserved residues might be important for functions that are unique to the Ca or Cb clades. This study also provides a good example of a seemingly simple phylogenetic problem which, due to a very high degree of sequence conservation and corresponding weak phylogenetic signals, combined with problematic nonphylogenetic signals, is nontrivial for state-of-the-art probabilistic phylogenetic methods. Citation: Søberg K, Jahnsen T, Rognes T, Skålhegg BS, Laerdahl JK (2013) Evolutionary Paths of the camp-dependent Protein Kinase (PKA) Catalytic Subunits. PLoS ONE 8(4): e doi: /journal.pone Editor: Narayanaswamy Srinivasan, Indian Institute of Science, India Received December 12, 2012; Accepted March 5, 2013; Published April 12, 2013 Copyright: ß 2013 Søberg et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: The project was funded by the Research Council of Norway. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * j.k.lardahl@medisin.uio.no Introduction Protein kinases are enzymes that catalyze the transfer of a phosphate group from adenosine 59-triphosphate (ATP) to a serine, threonine, tyrosine or other residue on a substrate. Most eukaryotic protein kinases derive from a common ancestor kinase, and share the same core catalytic domain [1]. PKA (EC ) is a serine/threonine kinase which is ubiquitously expressed in the human body. It is involved in many intracellular signaling events, and its function, specificity, and downstream effects depend on factors such as subcellular localization, expression of a number of isoforms and physiochemical features [2,3]. The inactive form of PKA is a heterotetrameric holoenzyme consisting of a regulatory (R) subunit dimer binding to two catalytic (C) subunits [4]. During activation, camp binds cooperatively to two sites termed A and B on each R subunit. In the inactive holoenzyme, only the B site is exposed and available for camp binding. When occupied, this enhances the binding of camp to the A site which leads to an intramolecular conformational change and the release of the R subunit dimer. The two C monomers are then free to phosphorylate relevant C substrates in the cytosol and nucleus [5 7]. Thus, a major function of the R subunit is to inhibit the phosphotransferase activity of the C subunits through direct interactions. Several variants of the C and R subunits have been identified in human cells. Four R subunits designated RIa, RIb, RIIa, and RIIb are transcribed from separate genes [8]. PKA holoenzymes containing RI and RII subunits are designated PKA type I and II, respectively [9,10]. Protein-protein interactions and organization of signal transduction pathways is required to obtain specificity in space and time. Protein kinases are localized to the relevant subcellular sites through anchoring, scaffolding and adapter protein activity. Major organizers of the camp signaling pathway are the A kinase anchoring proteins (AKAPs), which both function as scaffolding proteins and attach PKA to subcellular structures [11,12]. Initially, AKAPs were shown to interact with the RII subunits [13]. Later, dual specific AKAPs binding both RII and RI, as well as AKAPs binding only RI, have been identified, demonstrating that both PKA type I and II may be tethered to subcellular compartments in the cell [14 16]. Five different human C subunit genes have been identified; PRKACA, PRKACB, PRKACG, PRKX, and PRKY [2,17 19], all within the AGC group of kinases, which contains the cyclicnucleotide-dependent family (PKA and PKG), the protein kinase C family, the b-adrenergic receptor kinase, the ribosomal S6 family and some other relatives of these kinases [1,20]. Three of the human C subunit genes, PRKACA, PRKACB, and PRKX, have been demonstrated to be transcribed and translated into functional protein kinases, termed PKA Ca, PKA Cb, and PRKX, respectively. Ca exhibits two splice variants, Ca1 [21] and Ca2 [22 24], by employing two alternative 59 exons in PRKACA (Fig. 1A). Whereas Ca1 is ubiquitously expressed in man, Ca2 is exclusively expressed in the sperm cell and has been shown to be PLOS ONE 1 April 2013 Volume 8 Issue 4 e60935

2 essential for sperm motility and fertilization [24 27]. A number of isoforms of human PRKACB have been identified with alternative splicing of exons 59 of exon 2 (Fig. 1A), encoding at least the following proteins: Cb1, Cb2, Cb3, Cb4, Cb3ab, Cb3b, Cb3abc, Cb4ab, Cb4b, and Cb4abc [28 33]. In addition, Cb variants formed by skipping of exon 4 are expressed in the brain of higher primates. These C subunits bind the R subunit in a campindependent fashion [34]. PRKX, which is a protein kinase encoded from the X- chromosome, also binds RIa in a camp sensitive fashion [19]. PRKX and PRKY are 94% identical and have unknown functions [18,20]. PRKACG is a retroposon lacking introns and may be a pseudogene [35]. Whereas mrna from PRKACG is solely, but ubiquitously, transcribed in the testis [17], the corresponding protein Cc has never been identified. In vitro experiments on expressed Cc have revealed a functional kinase with variantspecific properties. The Cc protein kinase is, in contrast to Ca, and most probably Cb, not inhibited by the Protein Kinase Inhibitor (PKI), and it requires higher concentrations of camp for dissociation from PKA type I holoenzymes [36,37]. The in vivo function of PKA Cc, if the protein exists, remains to be elucidated. The PRKACA and PRKACB genes share identical positions and intron phases for all nine introns, and are very likely the result of a gene duplication event (Fig. 1B). Human PKA Ca1 and Cb1 have the same length (350 residues) and share 93% sequence identity. Many studies have elucidated the importance and function of a number of residues and short sequence segments in these kinases, and some of these are briefly summarized in Fig. 1C. The sequence identity between human PKA Ca vs. Cc and PRKX is, 82% and 54%, respectively, and between Ca/Cb and other human kinases below 50%. PKA C kinases in invertebrate metazoa have sequence identity with human Ca/Cb above 77% (see below), confirming that the PKA Ca/Cb-like kinases, not including PRKX, builds a unique and compact clade/family of kinases within the AGC-group. A previous study explored the evolution of the R subunits of PKA [38]. Based on a multiple sequence alignment (MSA) of the most conserved region in the R subunit sequences, the phosphatebinding cassette, the authors proposed a new classification of the R subunits in place of a classification based on physiochemical properties. They also identified a signature sequence that characterizes the R subunit genes, and identified type- and subtype-specific residues. The same focused analysis of the C subunits of PKA is to our knowledge lacking. In order to get a comprehensive overview of all the known PKA genes and obtain insight into the essential residues of these proteins, we have collected, compared and performed an extensive analysis of PKA C subunit sequences from a large number of bilaterian animal species. The homologous genes PRKACA and PRKACB constitute a unique clade of kinases and their protein products are the main sources of PKA activity in the cell. Therefore, we focused our analysis on the Ca/Cb-like kinases, including Cc, but not the more remotely related kinases PRKX and PRKY. The main focus of this study has been on elucidating the phylogeny of vertebrate and other chordate PKA Ca/Cb homologs, while orthologous sequences from mollusks and arthropods were included mainly to serve as outgroups in the phylogenetic analysis. We show that an ancestor C subunit was duplicated around the time of the evolution of the first vertebrate species, giving rise to the paralogous genes encoding Ca and Cb. Further analysis on the Ca/Cb homologs revealed signature sequences characteristic of the two paralogs. Comparison of Ca and Cb sequences gave insight into molecular differences and possible mechanisms that may determine functional differences between these two prototype protein kinases. Materials and Methods Sequences of PKA Ca/Cb Homologs Homologous sequences of human PKA Ca/Cb were obtained from the UniProt [39] and NCBI [40] database resources and from the Ensembl project [41] as described in Materials and Methods S1. Standard BLAST sequence searching algorithms were employed [42] and only sequences with sequence identity (protein level) compared with human PKA Ca or Cb above 75% were included in the dataset for further analysis. The final dataset comprised 41 sequences from placental mammals, including 18 from primates, 5 marsupial sequences, 35 sequences from nonmammalian vertebrates, and 15 from invertebrates (See Materials and Methods S1 and Table S1). Both nucleotide and protein sequences, one single splice variant for each gene, were stored in FASTA format for the total 96 sequences. 81 sequences appear to be full-length, comprising exons 2 to 10 and in addition a 59 exon chosen to correspond to human Ca1/Cb1 when possible. Of the incomplete sequences, 8 were missing only the 59 exon. In addition, one Petromyzon marinus sequence was missing exons 1, 7, 9, and 10, and the two sequences from Macropus eugenii were missing exons 1 and 9, and part of the 39 end, respectively. Finally, the four cartilaginous fish sequences only contain fragments of the full-length sequence. The sequences were aligned with MUSCLE [43] and the MSAs were viewed and edited with Jalview [44] (See Figure S1). All analysis was subsequently based on MSAs corresponding to exons 2 10 of human PKA Ca1/Cb1 (i.e. residues , Fig. 1B) unless otherwise stated, and columns of the MSAs containing gaps were deleted. PHYLIP and NEXUS format files were generated from the FASTA files with a dedicated Perl script. The best model for nucleotide evolution was determined with ModelTest 3.7 [45] in combination with PAUP* (D. L. Swofford, PAUP*, Sinauer Associates, Sunderland, MA) as chosen by the Akaike Information Criterion (AIC). The evolutionary model selected was the general time reversal (GTR) model with a discontinuous gamma distribution (C) for modeling rate heterogeneity over sites and a proportion of invariant sites (I), i.e. GTR+C+I with four rate categories. The best fitting sequence substitution model for the protein data was determined with ProtTest 2.4 [46], and was found to be, according to the AIC, LG+C+I [47]. Also the second best model, JTT+C+I [48], was tested for phylogenetic tree construction. Phylogenetic Analysis Bayesian inference of phylogeny was carried out with MrBayes [49,50] with default heating parameters (three heated Markov chain Monte Carlo chains and one cold) and priors. Two simultaneous and independent runs were carried out with sampling every 10 of 500 k generations until average standard deviation of split frequencies were below Branch lengths and majority rule consensus tree topologies were calculated after discarding a burn-in of 100 k generations after which stationarity had been reached. For all calculations presented, the final potential scale reduction factor (PSRF) was below for all parameters. Phylogenetic trees were also generated with the Maximum Likelihood (ML) method employing PhyML 3.0 [51] with default parameters. The robustness of each clade was estimated by a nonparametric bootstrap analysis with 1000 replicates. Gamma distribution parameters, the proportion of invariable sites, branch PLOS ONE 2 April 2013 Volume 8 Issue 4 e60935

3 Figure 1. Gene structure and overview of the human PKA catalytic subunits Ca and Cb encoded by the PRKACA and PRKACB genes, respectively. A Exons, introns and 39 and 59 untranslated regions (UTRs) of the PKA catalytic subunit genes are shown. The human Ca gene (PRKACA) is located at chromosome 19p13.1 (reverse strand) and has a length of approximately nucleotides (nt). Alternative transcription start sites give rise to two splice variants known as Ca1 and Ca2 (formerly known as CaS). Both splice variants comprise exons 2 to 10 and in addition a 59 exon 1 1 or 1 2 in Ca1 and Ca2, respectively. The human Cb gene (PRKACB) is located at chromosome 1p31.1 (forward strand), with a length of approximately nt. Alternative splicing of exons 1 1, 1 2, 1 3 and 1 4 give rise to the splice variants Cb1, Cb2, Cb3 and Cb4, respectively. In addition, three short exons, a, b and c, have been shown to be included in the transcript in various combinations. All known enzymatically active isoforms of Cb comprise exons 2 to 10. B Human PKA Ca1 consists of ten a-helix and nine b-strand secondary structure elements [76]. The figure (middle box) gives the location of a-helices (pink, A J) and b-strands (yellow, 1 9) relative to the ten exons and 351 encoded codons of the Ca1 isoform. The locations of the boundaries between exons are given on the upper line. The codons corresponding to the nine introns, as well as their PLOS ONE 3 April 2013 Volume 8 Issue 4 e60935

4 intron phases, are given on the lower line. The intron phase is defined as the position of the intron within a codon, with phase 0, 1, or 2 lying before the first base, after the first base, or after the second base, respectively. Human PKA Cb1 has the same length as Ca1, and the two proteins are differing at only 25 amino acid positions (92.9% sequence identity), strongly suggesting that the overall 3D structures, including secondary structure elements, are close to identical. The position and intron phases for the nine introns are also conserved between human Ca1 and Cb1. The sequence segment corresponding to exons 2 10, termed Core , is shown as a blue bar. C The function of selected important residues and motifs in human PKA catalytic subunits has previously been elucidated in the literature. All listed residues, and their numbering, are identical in Ca1 and Cb1, but the research describing these residues and motifs has mainly been performed on Ca. Numbering of amino acids is given for mature Ca1 and Cb1 (with N- terminal Met removed), both encoding 350 residues. Gly1 is found to be posttranslationally modified by myristoylation [77]. Ser10, Ser139 and Ser338 are well characterized phosphorylation sites [78,79]. The Gly-rich loop (Gly50 Gly55) plays an important role in phosphoryl transfer [80 82]. Lys72 and Asp184 are crucial for ATP and Mg 2+ binding in the active site [81] and the DFG motif (Asp184 Gly186) is conserved in most kinases. The conformation of the motif is critical for the functional state of the kinase [83,84]. Phe327 is the only residue outside of the kinase core binding to the adenine of ATP [82]. Trp196 is an essential residue for R subunit binding and phosphorylation of Thr197 is necessary for the enzyme to assume the active conformation, thereby facilitating catalysis as well as R subunit binding [5,85]. The hydrophobic P+1 motif (Gly200 Glu208) is important for the structure of the enzyme, as well as for substrate recognition [80,86]. Tyr247 competes with camp for R subunit binding [5]. doi: /journal.pone g001 lengths and GTR model parameters were all optimized by the ML algorithm from the data. ML phylogenetic trees were also inferred using RAxML [52], but due to the similarities of the resulting trees only the PhyML results are shown here. Dendroscope [53] was used for visualization of phylogenetic trees. Signature sequence logos were generated by applying WebLogo [54]. Calculations were carried out on the University of Oslo Titan computer cluster, mainly through the freely available Bioportal ( Protein structure illustrations were generated with PyMOL (W. L. DeLano, The PyMOL Molecular Graphics System, Version 1.3, Schrödinger, LLC). Ratios of nonsynonymous and synonymous substitution rates were calculated with the KaKs_Calculator [55]. Results and Discussion The PKA Ca/Cb Family is Highly Conserved in Chordates Vertebrate homologs of human PKA Ca and Cb were extracted from public databases, including Ensembl, UniProt, and database resources provided by the NCBI. New gene models were generated for several of the homologs by careful manual curation (See details in Materials and Methods S1). Homologs, in most cases full-length sequences, were found from 21 placental mammals, the marsupial species opossum (Monodelphis domestica) and wallaby (Macropus eugenii), chicken, zebra finch (Taeniopygia guttata), the Carolina anole lizard (Anolis carolinensis), and two and six species of frogs and bony fishes, respectively. In addition, fulllength sequences were obtained for the non-vertebrate chordates amphioxus (Branchiostoma floridae) and the tunicates Ciona intestinalis and Ciona savignyi, the echinoderm sea urchin (Strongylocentrotus purpuratus), the two mollusks great pond snail (Lymnaea stagnalis) and California sea hare (Aplysia californica), the hemichordate Saccoglossus kowalevskii, the nematode Caenorhabditis elegans, the sponge Amphimedon queenslandica, as well as six arthropods including honey bee, fruit fly, a tick, and a crustacean. Partial sequences were obtained for orthologs from dogfish shark (Squalus acanthias), little skate (Leucoraja erinacea) and sea lamprey (Petromyzon marinus). In all cases both the nucleotide sequences and the corresponding protein sequences were stored. All sequences are listed in Materials and Methods S1. We were unable to detect more than a single, reasonably close, PKA Ca/Cb homolog in any of the invertebrate species that were examined. These includes the cephalochordate amphioxus, the urochordates C. intestinalis, C. savignyi and Oikopleura dioica, the echinoderm sea urchin, as well as arthropods, mollusks, a nematode and a sponge. Two homologs were found in most vertebrates, including sea lamprey, and the chondrichthyes S. acanthias and L. erinacea, while four homologs were detected in a number of bony fishes. The main isoforms of human PRKACA and PRKACB comprises exons 2 10 and in addition one or more 59 exons (Fig. 1A). The length of all exons 2 10 and the positions and intron phases of all nine introns are identical in the two human genes (Fig. 1B), clearly demonstrating that these genes arose due to a gene duplication. This structure, reflecting an extreme degree of conservation, is also conserved in all chordate homologs, including the invertebrate cephalochordate amphioxus and the urochordate tunicates and in addition in the echinoderm sea urchin, the hemichordate S. kowalevskii, and the crustacean Daphnia pulex: the number of exons, their lengths and the codon phases of all introns are identical. The exceptions, apart from the intronless homologs described below, includes the medaka gene encoding a protein with Ensembl identifier ENSORLP This gene appears to have acquired a new 88-nucleotide intron (intron phase 0) in the middle of exon 9. In addition, the basal metazoan A. queenslandica has an additional intron of 334 nucleotides that splits the coding sequence corresponding to vertebrate exon 6. The number of codons encoded by exons 2 10 in the PRKACA/ B family of genes - corresponding to residues in human proteins PKA Ca1/Cb1 - is consequently also identical in all species listed above. We term this sequence segment Core (Fig. 1B) in order to distinguish it from the variable 59 exons. The genomic data for the mollusks are not yet available in the public domain and the intron/exon structure is currently unknown, but in insects the protein coding segment of the PRKACA/B homologs are contained in a single exon. Nevertheless, the insect homologs also have the same number of residues for the segment corresponding to Core An MSA of all sequences with fulllength Core is shown in Figure S1. Consequently, we find that nearly all our PKA Ca/Cb homologs from Bilateria have the same length for the Core , and with variable length N-termini corresponding to the multitude of isoforms. The only exceptions are a single-residue insertion after human PKA Ca1 Lys63, in the middle of exon 3, in both homologs from mollusks and a singleresidue insertion after human PKA Ca1 Ala38 in a fast-evolving sequence from opossum (identifier ENSMODP ). These insertions are in loop structures in PKA Ca1 (See e.g. [56] for the 3D structure) and are expected to be compatible with an unchanged overall 3D structure of the kinase. Pairwise sequence identity between any of the bilaterian PKA Ca/Cb homologs described above is always above 77% for the 335 residues of the Core at the amino acid level. Leaving out the fast-evolving sequences from marsupials and a single fastevolving zebrafish sequence (Q7T374), the sequence identities are always above 80% and 87% within the chordates and vertebrates, respectively. This demonstrates very strong purifying selection and an exceptionally high degree of sequence conservation for this protein family. PLOS ONE 4 April 2013 Volume 8 Issue 4 e60935

5 The PKA Ca/Cb Gene Family Contains Several Putative Retroposons In addition to the PRKACA/B homologs that have conserved exon/intron structure in chordates and several other deuterostomes, and the arthropod homologs, also with several exons, but with the protein coding segment contained in a single exon, a number of intronless PRKACA/B homologs are found in vertebrate genomes. Among these are human PRKACG, located on chromosome 9 between the genes PIP5K1B and FXN, but transcribed in the opposite direction. Intronless PRKACG has the same number of codons as the PKA Ca1-like transcript of PRKACA and is conserved in the great apes, in chimpanzee, gorilla, and orangutan and in the Old World monkeys rhesus macaque and hamadryas baboon (Sequences are listed in Materials and Methods S1). Genome browsing at the Ensembl resource also shows the synteny to be conserved in these species with PIP5K1B and FXN being transcribed in one direction and PRKACG, located between these two genes, in the opposite. Between PIP5K1B and FXN in the gibbon (Nomascus leucogenys) genome, a species more closely related to great apes than the Old World monkeys, there is no full-length PRKACG ortholog, but instead a putative PRKACA/B pseudogene with several frame shifting mutations. We find no evidence of PRKACG orthologs in prosimian primates such as greater galago (Otolemur garnettii), tarsier (Tarsius syrichta), or gray mouse lemur (Microcebus murinus), although the last two of these have genomes that are still fragmented and with undisclosed synteny around the genes PIP5K1B and FXN. In the common marmoset (Callithrix jacchus) genome, there is a fragment, most likely not protein-coding, of PRKACG between PIP5K1B and FXN, corresponding to PKA Ca1 residues Finally, in the mouse and dog genomes, PIP5K1B and FXN are neighboring genes being transcribed in the same direction, but without any sign of a PRKACA/B homolog in this region. These findings strongly support the previous suggestion that PRKACG is a retroposon due to a PKA Ca1-type transcript [35] that has been inserted between PIP5K1B and FXN in a common ancestor of great apes and Old and New World monkeys. The putative PRKACG transcripts could potentially give rise to functional kinases in all great apes and in the Old World monkeys, but not in gibbons and the New World monkey marmoset where there are mutations disrupting the reading frame in the PRKACG retroposon. In the marmoset genome, there is in addition to the putative PRKACG pseudogene on chromosome 1, a second retroposon (Ensembl identifier ENSCJAP ) related to PKA Ca1 on chromosome 2. This gene was not found in other primates. In addition to the PKA Ca1-like retroposons in primates, we found intronless PRKACA homologs in the two sequenced genomes of marsupials, the wallaby kangaroo (M. eugenii) and the Brazilian opossum (M. domestica) (See Materials and Methods S1). Also these putative retroposons appear to be derived from a PKA Ca1-type transcript, but are otherwise unrelated to primate PRKACG. Elucidating the Phylogeny of the PKA Ca/Cb Family is Nontrivial Due to High Sequence Conservation The distribution of PKA Ca/Cb homologs in Bilateria, as well as phylogenetic trees generated with unsophisticated hierarchical clustering methods (results not shown), suggest that this gene family has expanded through repeated gene duplication events in vertebrates. However, despite numerous attempts, state-of-the-art probabilistic methods, both Bayesian inference and maximum likelihood (ML) methods, were not able to generate a statistically strongly supported phylogenetic tree for the full PKA Ca/Cb family from the complete data set. After careful analysis of the data (vide infra), we were nevertheless able to derive reliable phylogenies for this gene family by dividing the data into subsets. The final phylogenetic trees were generated as follows: an MSA was generated from the nucleotide sequences corresponding to the Core segment for selected vertebrates, amphioxus, sea urchin and fruit fly. After removal of all nucleotides at codon position 3, the dataset was employed to generate Bayesian inference and ML trees, with identical topology, for the PKA Ca/Cb homologs with good Bayesian posterior probabilities and bootstrap support for the major nodes. The phylogram was rooted with the sea urchin and fruit fly as outgroups (Fig. 2). Bayesian inference methods were also used to generate a phylogenetic tree of 22 vertebrate PKA Ca orthologs employing human and mouse PKA Cb as outgroups (Fig. 3A). Similarly, a tree with 26 PKA Cb orthologs was generated with human and mouse PKA Ca as outgroups (Fig. 3B). These trees, and trees from the corresponding ML analysis, are based on MSAs for the nucleotide sequences with all three codon positions included. Neither Bayesian inference nor ML methods, generally accepted to be the most accurate [57,58], were able to generate a reliable phylogeny for the PKA Ca/Cb family from the full data set. This was not due to unreliable MSAs as only three of the sequences in the original data set each had single codon insertions (vide supra) and manual removal of these was trivial. The problem, however, appears to be a combination of weak phylogenetic signals and problematic nonphylogenetic signals in the data. Due to the very high level of sequence conservation, the protein data set contains few phylogenetically informative sites, i.e. amino acid sites that favor one phylogenetic tree topology over others. As an example, for the 22 chordate taxa used to generate the tree in Fig. 2, the corresponding amino acid data set has 335 columns/ sites. Of these, only 58 sites are phylogenetically informative, while 255 are fully conserved and identical for all taxa and the remaining 22 are autapomorphic. Unsurprisingly, protein data phylogenetic trees generated with two recommended substitution models (LG/JTT+C+I) and well-tested ML programs (PhyML/ RAxML) were overall fairly similar, but with very poor bootstrap support. The trees generated with various subsets of the available taxa were incongruent and also in several cases inconsistent with the known evolutionary relationship between the various chordate species. In order to secure a stronger phylogenetic signal [59], the nucleotide data set was employed for deriving the phylogenetic relations. For the 22 chordate taxa in Fig. 2, there are 77, 33, and 312 phylogenetically informative sites (out of 335 sites in total) for codon position 1, 2, and 3, respectively. The low number of informative sites at codon position 1 and 2 reflects the high degree of conservation at the amino acid level, while the high number at codon position 3 (93%) reflects the large time-span since the common ancestor of these genes. For all tested nucleotide data sets, the GTR+C+I model was predicted to be superior. While using the nucleotide sequences for deriving the phylogenetic relationships ensures a stronger phylogenetic signal, in particular the data from codon position 3, due to the degeneracy of the genetic code, may be severely mutationally saturated due to reversions and convergences that erase the true phylogenetic signal. This will especially be problematic for inferring ancient phylogenies [59], where long branch attraction (LBA), i.e. a tendency for grouping of lineages with long branches irrespective of their true relationships, may lead to misleading phylogenies [60]. In particular, fast-evolving genes may artificially occur too deeply in the tree due to LBA towards the outgroups. PLOS ONE 5 April 2013 Volume 8 Issue 4 e60935

6 Figure 2. Phylogenetic relationships among the PKA catalytic subunit homologs in chordates. The Ca and Cb paralogs are a result of a gene duplication in a common ancestor of vertebrates. Subsequent duplications of Ca and Cb in a teleost fish ancestor have resulted in four PKA catalytic subunits in these organisms. The Bayesian inference tree is based on the nucleotide sequences (codon positions 1 and 2 only, GTR+C+I model) of exons 2 to 10 which corresponds to a multiple sequence alignment with no gaps. The phylogram is shown with estimated branch lengths proportional to the number of substitutions at each site, as indicated by the scale bar. The arthropod fruit fly (D. melanogaster) and the echinoderm sea urchin (S. purpuratus) have been set as outgroups. Bayesian posterior probabilities are shown for each node. The topology of a maximum likelihood (ML) tree generated with the same data set and model was identical to the Bayesian inference tree. ML bootstrap values are shown for selected nodes (1000 replications). The sequences of human and mouse PKA Ca and Cb and the homologs from amphioxus (B. floridae), zebra finch (T. guttata), chicken (G. gallus), the frog X. tropicalis, the lizard A. carolinensis, medaka (O. latipes), the pufferfish T. rubripes, and stickleback (G. aculeatus) are described in Materials and Methods S1. The X. tropicalis Ca and A. carolinensis Cb are incorrectly placed (See discussion and Fig. 3). doi: /journal.pone g002 As for the protein data set, trees generated with subsets of the available taxa were highly incongruent and in several cases with pronounced LBA, clearly demonstrating a strong nonphylogenetic signal. In order to secure minimal nonphylogenetic signals and reduce LBA, codon position 3 sites were removed from the data set used to generate the phylogeny in Fig. 2. In addition, fast-evolving taxa such as the tunicate [61], marsupial PKA Ca/Cb homologs and the primate PKA Cc homologs were left out of the data set. Fig. 2 is expected to describe the ancient evolution of the PKA Ca/Cb family in chordates correctly, with the gene duplications in the common ancestor of vertebrates as well as in teleost fishes supported by high Bayesian posterior probabilities and strong bootstrap support for the ML analysis. In order to better describe the evolution of the two PKA Ca and PKA Cb paralogs in vertebrates separately, the trees in Fig. 3 were generated. Reliable phylogenies could not be generated from a data set containing codon positions 1 and 2 only, as in Fig. 2, most likely due to weak phylogenetic signals in the data. However, for the relatively recent phylogenetic relations in Fig. 3, the saturation at codon position 3 is expected to be less severe and all three codon positions were included in the data set for analysis. In Fig. 2, there are two errors due to nonphylogenetic signal. These are the placement of frog PKA Ca in a clade together with teleost fish and the erroneous lizard/frog PKA Cb clade. Both these errors are corrected in Fig. 3 upon inclusion of codon position 3 data. The most ancient splittings, however, are unreliable in Fig. 3, in particular the description of the branching between the tetrapod and teleost fish PKA Ca homologs as a multifurcation (Bayesian analysis) or two bipartitions with extremely poor bootstrap support (ML analysis, not shown) in Fig. 3A. In conclusion, while the phylogenies of the PKA Ca and PKA Cb subfamilies, especially within the tetrapods, appears to be correctly inferred from the full nucleotide dataset (Fig. 3), the ancient evolution of the PKA Ca/Cb family in chordates is correctly described with cladistic methods only after removal of codon position 3 data (Fig. 2). The protein data set contains PLOS ONE 6 April 2013 Volume 8 Issue 4 e60935

7 Figure 3. The Bayesian inference trees for vertebrate PKA Ca and Cb both closely reflects the evolutionary relationships among these organisms. A Phylogenetic analysis of Ca orthologs resulted in a tree that was rooted with human and mouse Cb as outgroups. The tree was based on the nucleotide sequences of exons 2 to 8 (all codon positions, GTR+C+I model). B Phylogenetic analysis of Cb orthologs was performed PLOS ONE 7 April 2013 Volume 8 Issue 4 e60935

8 employing nucleotide sequence data (all codon positions, exons 2 to 10, GTR+C+I model). The resulting tree was rooted with human and mouse Ca as outgroups. In both trees, branch lengths are shown as substitutions per site, with scale indicated by the scale bars. Bayesian posterior probabilities are given for each node and ML bootstrap values (1000 replications) are shown for selected nodes where the clades are identical in the Bayesian and ML analysis. In addition to organisms found in Fig. 2, representative sequences from the following species were included: eutherian mammals rhesus macaque (M. mulatta), tarsier (T. syrichta), dog (C. familiaris), horse (E. caballus), pig (S. scrofa), cow (B. taurus), rat (R. norvegicus), and hamster (C. griseus), marsupial mammals wallaby (M. eugenii) and opossum (M. domestica), the frog X. laevis, the pufferfish T. nigroviridis and Atlantic salmon (S. salar). See Materials and Methods S1 for the sequence data. doi: /journal.pone g003 limited phylogenetic information and does not give reliable phylogenies in chordates. No analysis based on partitioned data, for example according to codon position, was attempted, as this is likely to lead to overparametrization. Phylogenetic Inferences for the PKA Ca/Cb Family in the Chordate Lineage The well-resolved phylogenetic tree in Fig. 2 strongly suggests the following sequence of events during the evolution of the PKA Ca/Cb gene family: a single PKA Ca/Cb-like gene in the common ancestor of chordates, arthropods and echinoderms was duplicated in a common ancestor of vertebrates, which lead to the two paralogous genes corresponding to PKA Ca and Cb. These two genes were again duplicated in a common ancestor of teleost fishes, leading to paralogs that we suggest are denoted PKA Ca-I, Ca-II, Cb-I, and Cb-II. The data indicates that the first PKA Ca/Cb gene duplication, resulting in PKA Ca and PKA Cb, took place after the divergence of the urochordate and cephalochordate lineages. Currently there are only fragments of PKA Ca/Cb homologs available in public databases for the chondrichthyes (dogfish shark and little skate) and the cyclostome sea lamprey (Table S1), but the PKA Ca/Cb homologs also in these organisms appear to occur in pairs. Unfortunately, the phylogenetic signal in the data is too weak to classify these sequences as PKA Ca or PKA Cb, and to exclude the possibility that these paralogous gene pairs are results of independent gene duplications, but the most parsimonious explanation for this distribution of PKA Ca/Cb homologs is that a single PKA Ca/Cb gene duplication occurred before the divergence of the jawless fish lineage and the subsequent divergence of sharks and skates. Interestingly, this timing of the PKA Ca/Cb gene duplication coincides with the two rounds (2R) of whole genome duplication (2R hypothesis) that took place after the emergence of the invertebrate chordates and before the radiation of jawed vertebrates [62 64]. The PKA Ca and PKA Cb gene split is thus likely to have occurred in the Cambrian, roughly 500 Mya [65,66]. Canaves and Taylor [38] found that the gene duplications of the PKA regulatory subunits resulting in the paralogs RIa and RIb as well as RIIa and RIIb also occurred in the chordate lineage, suggesting that the gene duplications of the R and C subunits might have taken place simultaneously. The secondary duplications of PKA Ca and PKA Cb (Figs. 2 and 3) appear to be unique to teleost fishes, and might have coincided with the teleost whole genome duplication that took place Mya [67,68]. Finally, the presence of two X. laevis PKA Ca paralogs and a single X. tropicalis PKA Ca (Fig. 3A) in the amphibian genomes is consistent with the recent whole genome duplication event in the common ancestor of the X. laevis group not found in X. tropicalis [69,70]. As expected, both the subtrees for PKA Ca (Fig. 3A) and PKA Cb (Fig. 3B) have eutherian clades with the marsupial orthologs as sister clades, and with Sauropsida (birds and reptiles, Fig. 3B only) and frogs appearing as sister clades deeper into the phylogenetic tree. We were unable to find the PKA Ca gene in any of the released genomes of Sauropsida, and this clade is consequently missing in Fig. 3A. However, a single EST (expressed sequence tag) sequence from a chicken testis library (GenBank identifier CN [71]) appears to confirm the presence of PKA Ca also in birds. Several vertebrate species are missing either PKA Ca or PKA Cb in the current genomic data sets, but this is most likely due to low sequence coverage in unfinished genomes. We find no evidence for extensive PKA Ca or Cb gene loss in any of the main vertebrate groups. The marsupial M. domestica PKA Cb appears to be particularly fast-evolving (Fig. 3B). The M. eugenii PKA Cb ortholog is present in the genome, but the full sequence is currently unknown. Both marsupial genomes [72,73] have two PKA Ca paralogs. Marsupial PKA Ca-I has the same exon/intron structure as PKA Ca in all other mammals. Marsupial PKA Ca-II, however, is intronless and a putative PKA Ca retroposon (Fig. 3A). The full-length sequences are not at present available for all the marsupial PKA Ca homologs, but interestingly, intronless PKA Ca-II appears to be significantly more conserved than the ancestral variant PKA Ca-I. The 59 segment of the two marsupial PKA Ca-II orthologs (only the codons are available in the sequence for M. eugenii) have 17 synonymous and no non-synonymous mutations, suggesting strong purifying selection. For PKA Ca-I, the 294 codons corresponding to exons 2 9 (exons 1 and 10 are missing in the current M. eugenii genomic sequence) have 70 mutations, including a 3 nucleotide insertion in opossum PKA Ca-I, and 21 of these are non-synonymous. These data, although limited, strongly suggests that the retroposon PKA Ca-II have become functional in marsupials and that the purifying selection acting upon PKA Ca-I and PKA Cb has become less stringent. Vertebrate PKA Ca and Cb Mainly Differs in the C-tail and in Subdomains I and II In order to elucidate the potential differences between the PKA Ca and PKA Cb protein subfamilies, an MSA of all available vertebrate homologs was generated. In this set of 27 PKA Ca and 33 PKA Cb, there are within the 335 sites/columns of the Core only 62 sites (19%), that are phylogenetically informative while 235 sites are fully conserved in all taxa, again reflecting the very high degree of purifying selection. A manual inspection of the MSA showed that eleven sites/columns could tentatively be used to discriminate between the two PKA subfamilies. Sequence logos for these eleven sites are shown in Fig. 4. Human PKA Ca1 residues Gln35, Thr37, Glu64, Gly66, His68, Ser109, and Glu334 are fully conserved in all vertebrate PKA Ca, while at the corresponding sites in PKA Cb the sequence conservation is less stringent. Similarly, human PKA Cb1 residues Asp42, Gln67, Arg319 are fully conserved in vertebrate PKA Cb (Fig. 4). The single site where there is no overlap between amino acid use in the two subfamilies is at residues 66 where PKA Ca and Cb have Gly and Glu/Asn/Asp, respectively. The residues that correspond to the eleven sites that tentatively discriminates between the subfamilies PKA Ca and Cb are all exposed at the protein surface, in or close to loop structures, in subdomains I and II and in the C-tail (Fig. 5). None of these residues are located close to the kinase active site and their identity PLOS ONE 8 April 2013 Volume 8 Issue 4 e60935

9 Figure 4. The identity of eleven amino acids in the protein chain may define the Ca and Cb branches of PKA catalytic subunits. Our full set of PKA catalytic subunits (Materials and Methods S1) from bony fishes and tetrapods, comprising 27 Ca and 33 Cb, was employed to identify eleven amino acid positions that together may be used to classify a PKA catalytic subunit as belonging to one of the two branches. The sequence logos define the PKA Ca and Cb clades within the Teleostomi, which includes the familiar classes of bony fishes, birds, mammals, reptiles, and amphibians. We find invariable Gln35, Thr37, Glu64, Gly66, His68, Ser109 and Glu334 in Ca and invariable Asp42, Gln67, and Arg319 in Cb (Ca1/Cb1 numbering). The residues in the corresponding positions in human Ca1 and Cb1 are also shown. doi: /journal.pone g004 is not likely to affect PKA kinase activity. Likewise, they are not located near the protein surface segments that are known to interact with the PKA regulatory subunits (Fig. 5B), and these residues should not be important for R subunit interactions. Several of the eleven residues, especially residues 64, 319, 334 and 340, are protruding their side chains into the solvent and might be targets for PKA Ca- or PKA Cb-specific post-translational modifications and/or protein-protein interactions. Particularly, Glu64 is conserved in all vertebrate PKA Ca, while residue 64 is variable in 33 Cb, being Ala in all mammals and Glu in only three fish homologs. In PKA Cb Glu66 is absolutely conserved, except in five of the fish homologs, while this residue is conserved as Gly in Ca. Strong Purifying Selection is Lost in PKA Cc A comparison of PKA Ca1 from human and the galago, a prosimian primate, shows that there are 59 and 4 synonymous (silent) and non-synonymous (amino acid changing) mutations, respectively. Between human and macaque PKA Cc there are slightly fewer, 44, mutations, but from these 31 amino acid changes, indicating that the strong purifying selection in the PKA Ca lineage is lost in PKA Cc. A powerful tool for evaluating the evolution of protein coding sequences, is calculating the ratio of the non-synonymous (Ka) and synonymous (Ks) substitution rates, where Ks is the number of synonymous substitutions per synonymous site and Ka is the number of non-synonymous substitutions per non-synonymous site [74]. A Ka/Ks,1 indicates purifying (negative) selection, Ka/Ks.1 is a sign of positive selection, while Ka/Ks, 1 indicates neutral evolution of the protein. Table 1 shows Ka/Ks for comparisons of five representative tetrapod PKA Ca homologs and five primate PKA Cc homologs. Due to missing sequences in the databases it was not possible to compare the same species for Ca and Cc. For the PKA Ca sequences, the average Ka/Ks is PKA Ca sequences were also compared for other placental mammals and Ka/Ks were found to be in the range Ka/Ks for a comparison of human PKA Ca and Cb, and of the human and mouse PKA Cb orthologs, were and , respectively. These data for the PKA Ca/Cb homologs in placental mammals again confirms the very strong purifying selection acting upon these kinases. The Ka/Ks values from the comparison of the primate PKA Cc sequences in Table 1 are in the range Due to the fairly close evolutionary relationships between these species, the number of mutations in PRKACG transcripts between these species is rather low, for example 9, 31 and 46 between human and chimpanzee, orangutan, and macaque, respectively. Consequently, the Ka/Ks values are not expected to be highly reliable, but the average value of 0.45 clearly suggests that the strong purifying selection in the PKA Ca lineage is lost in PKA Cc. This finding that mutations in PKA Cc appear to be neutral, combined with the loss of a functional Cc in gibbons and marmoset (vide supra), suggest that there are no evolutionary constraints on maintaining a functional PKA Cc protein in higher primates. However, this analysis does PLOS ONE 9 April 2013 Volume 8 Issue 4 e60935

10 Figure 5. Signature residues defining PKA Ca and Cb do not interact with ATP, peptide inhibitor PKIa or the kinase regulatory subunit. A The tentative signature residues of PKA Ca and Cb (Fig. 4) are highlighted in a structural model of PKA Ca1 in complex with a truncated PKIa (residues 6 25). Signature amino acids in human Ca1 and Cb1 are shown without and within parenthesis, respectively. The conserved kinase core has been divided into subdomains (represented in different colors) [87] as defined by Hanks and Hunter [88]. ATP is rendered as sticks (black) and two divalent cations as black spheres. The model is based on the experimental structure of Thompson et al. [56] (PDB identifier 3FJQ). B Residues in Ca1 (cyan) interacting with regulatory subunit RIa (purple, residues of bovine RIa only) are mainly restricted to the large lobe and do not overlap with any of the Ca signature residues (red). The complex is shown with the same orientation of Ca1 as in panel A (left) and rotated 180u (right). The model is based on the experimental structure of Kim et al. [5] (PDB identifier 3FHI). doi: /journal.pone g005 PLOS ONE 10 April 2013 Volume 8 Issue 4 e60935

11 Table 1. Ratio of amino acid replacing (Ka) and silent (Ks) mutation rates for all pairwise comparisons of five tetrapod PKA Ca homologs and five primate PKA Cc homologs. Ka/Ks a H. sapiens Ca O. garnettii Ca M. musculus Ca B. taurus Ca O. garnettii Ca (0.0158) M. musculus Ca (0.0141) (0.0226) B. taurus Ca (0.0063) (0.0111) (0.0158) X. tropicalis Ca (0.0060) (0.0056) (0.0064) (0.0060) Ka/Ks b H. sapiens Cc P. troglodytes Cc P. pygmaeus Cc G. gorilla Cc P. troglodytes Cc (0.127) P. pygmaeus Cc (0.324) (0.194) G. gorilla Cc (1.207) (0.148) (0.319) M. mulatta Cc (0.678) (0.406) (0.393) (0.603) a Ka/Ks ratios for pairwise comparisons of PKA Ca from human, a prosimian primate (O. garnettii), mouse (M. musculus), cattle (B. taurus) and a frog (X. tropicalis) calculated according to the model of Goldman and Yang [89] and the model averaging method of Zhang et al. (in parenthesis) [90] for the Core sequence segment. b Ka/Ks ratios for pairwise comparisons of PKA Cc from human, chimpanzee (P. troglodytes), orangutan (P. pygmaeus), gorilla (G. gorilla) and rhesus macaque (M. mulatta) calculated as described above. doi: /journal.pone t001 not exclude the possibility that the PKA Cc transcript has an important function in humans, great apes and other Simiiformes, for example in regulation of PKA Ca/Cb transcript processing [75]. Conclusion We have shown that the PKA Ca and Cb catalytic subunits found in chordates and other animal species builds a phylogenetic clade of kinases with a very high degree of conservation at the protein level. In the core segment corresponding to exons 2 10 of vertebrate Ca1/Cb1 the synonymous mutation rate is approximately two orders of magnitude larger than the amino acid changing mutation rate. All the main residues and sequence segments previously shown to be important for human Ca/Cb function (Fig. 1C), including the phosphorylation sites, the ATP and Mg 2+ interacting residues and the DFG and P+1 motifs, are basically fully conserved in all homologs in chordates, insects and other animal sequences investigated in the current study. The few residues that differ in Ca and Cb (Fig. 4) should be investigated in order to elucidate possible functional differences between the two paralogs. Finally, the Ca1-derived expressed retroposon Cc found in higher primates appears to be evolving neutrally and appears to have no function as a mature protein. Supporting Information Figure S1 Multiple sequence alignment of segment corresponding to exons 2 10 of human PKA Ca1 (i.e. References 1. Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S (2002) The protein kinase complement of the human genome. Science 298: Skålhegg BS, Taskén K (2000) Specificity in the camp/pka signaling pathway. Differential expression, regulation, and subcellular localization of subunits of PKA. Front Biosci 5: D678 D Taylor SS, Yang J, Wu J, Haste NM, Radzio-Andzelm E, et al. (2004) PKA: a portrait of protein kinase dynamics. Biochim Biophys Acta 1697: residues ). All sequences are described in Materials and Methods S1. Residue numbering of human PKA Ca1 (identifier P17612) is shown above the sequence. Sequences belonging to the Ca, Cc, and Cb clades are marked at the right by a green, blue, and red bar, respectively. The figure was prepared with Jalview [44]. (PDF) Table S1 Sequence data for PKA catalytic subunit homologs from chordates collected and manipulated as described in Materials and Methods S1. (PDF) Materials and Methods S1 collection. (PDF) Acknowledgments Details on sequence data We are grateful to Russell Orr and Kamran Shalchian-Tabrizi for help and useful comments and the University of Oslo Bioportal for computational infrastructure. Author Contributions Conceived and designed the experiments: KS TJ TR BSS JKL. Performed the experiments: KS JKL. Analyzed the data: KS TJ TR BSS JKL. Contributed reagents/materials/analysis tools: KS JKL. Wrote the paper: KS TJ TR BSS JKL. 4. Krebs EG, Beavo JA (1979) Phosphorylation-dephosphorylation of enzymes. Annu Rev Biochem 48: Kim C, Xuong NH, Taylor SS (2005) Crystal structure of a complex between the catalytic and regulatory (RIa) subunits of PKA. Science 307: Anand GS, Krishnamurthy S, Bishnoi T, Kornev A, Taylor SS, et al. (2010) Cyclic AMP- and (R p )-camps-induced conformational changes in a complex of the catalytic and regulatory (RIa) subunits of cyclic AMP-dependent protein kinase. Mol Cell Proteomics 9: PLOS ONE 11 April 2013 Volume 8 Issue 4 e60935